Files
ALwrity/docs-site/docs/user-journeys/developers/podcast-maker-journey.md

73 lines
2.8 KiB
Markdown

# Podcast Maker Journey - Developers
Use this journey to integrate Podcast Maker into repeatable, testable pipelines for scripted audio generation and distribution.
## Overview
### Entry Conditions
- **Inputs:** API credentials, topic payload schema, content constraints, output destination.
- **Skill level:** Intermediate to advanced (API and workflow automation).
- **Expected time:** 60-120 minutes for first implementation.
### Success Target
Automate one full podcast generation path from prompt to exported artifact with predictable quality.
## Setup
### Recommended Defaults
- **Duration:** 10-20 minutes (configurable per template)
- **Speakers:** 1-2 synthetic speakers
- **Voice style:** Neutral/professional with stable pacing
- **Research provider:** Perplexity (structured fact gathering for scripted outputs)
### Pre-Production Checklist
1. Define request schema for analysis/research/script/render/export stages.
2. Store provider credentials via environment variables.
3. Configure retry/error policy for external research and render calls.
4. Add logging for prompt versions and output hashes.
## Production
### Podcast Maker Workflow
1. **Analysis**
- Validate input payload and enforce required fields.
- Derive episode objective and section plan programmatically.
2. **Research**
- Fetch source context with provider abstraction.
- Normalize citations and drop low-confidence results.
3. **Script**
- Generate structured script JSON (intro/segments/outro/CTA).
- Run lint-style checks for length and forbidden terms.
4. **Render**
- Render audio using configured speaker profile.
- Execute post-render QA hooks (duration, loudness, clipping checks).
5. **Export**
- Persist artifact + metadata to storage.
- Trigger downstream publish/webhook integration.
## Optimization
### Success Criteria
- End-to-end pipeline completes without manual intervention.
- Output passes automated quality checks.
- Metadata includes provenance for research and prompt version.
- Failure paths are observable with actionable logs.
### Checkpoints
- **Before render:** Unit/integration checks pass for script payload.
- **After render:** Verify duration bounds and transcript alignment.
- **After publish:** Monitor error rate, latency, and output quality metrics.
## Troubleshooting
### Common Issues and Fixes
- **Provider timeouts:** Add retries with exponential backoff and fallback provider.
- **Inconsistent scripts:** Pin model settings and enforce schema validation.
- **Audio quality failures:** Add deterministic render settings and QA thresholds.
- **Broken exports:** Validate storage credentials and file naming conventions.
- **Debug difficulty:** Log stage-level inputs/outputs with correlation IDs.
---
Next step: integrate this into **[Advanced Usage](advanced-usage.md)** automation patterns.